Canon XC10
Transcript of Canon XC10
Canon XC10 July 2015
1
EBU - Tech 3335 : Methods of measuring the imaging performance of
television cameras for the purposes of characterisation and setting
Alan Roberts, July 2015
SUPPLEMENT 17: Assessment of a Canon XC10 camera
Tests have been conducted in line with EBU R.118. This document is a report of the results of
the tests defined in Tech3335 and is not an endorsement of the product.
This is a report on tests carried out on a Canon XC10 camera, serial number 953054000076, a European
version. It closely resembles a small DSLR rather than a conventional television camera, but has an integral
lens and no monocular viewfinder. However, a monocular viewfinder can be simulated by using an
accessory mirror-box adaptor on the LCD touch-screen display, but this prevents access to some control
functions which are operated only on the touch-screen.
According to the specification, the sensor is a single large-format CMOS with the Bayer pattern of photo-
sites; 8.29Meg for video (3840x2160), 12Meg for stills (4,000x3,000), and nominally 1” size. Further
information from Canon reveals that the full sensor is 4224x3164, actually 16.8mm diagonal. Thus the full
sensor dimensions are 13.48x10.10mm and the photo-sites are spaced at 3.192 microns, giving a photo-site
area of 10.19 square microns, about 40% of that of a conventional ⅔” camera. The video image (both 4k and
HD) is made from a central 3840x2160 part which must be 12.26x6.89mm and 14.1mm diagonal. The
effective photo-site size for HD is 6.38 microns square, similar to that of a conventional ⅔” camera.
The lens is fixed (not inter-changeable) with a 10:1 zoom range from 27.3 to 273mm (35mm equivalent).
Maximum aperture is F/2.8 and ramps to about F/5.6 at the long end of the zoom. It has a manual zoom ring
and a focus ring which is disabled in ‘auto-focus’ mode, so is not calibrated.
All recording is 8-bit MPEG4 AVC/H.264 4:2:2. HDTV recording is onto SDHC card (long-GoP, at 50Mb/s
for 50P, 35mB/s for 25P or 25I), 4k recording is onto CFast card (3840x2160 intra-frame at 305Mb/s for
50P, 205Mb/s for 25P). Recording can also be done ‘off-speed’ between 4 and 1/1200 times normal, and
there’s a 5-second cache for pre-recording (only in HD).
The camera weighs 1kg with the internal battery (7.2v) and consumes 6.2W when recording 4k, 5.4W when
recording HD. The removable battery is 1.865Ah (14Wh) so should last at about 2.5 hours in normal video
use. The battery can be charged in the camera or in an external charger.
There are connectors for HDMI output (disabled when Wi-Fi is enabled), USB for file transfer, 3.5mm
microphone, 3.5mm headphone.
Controls include three assignable buttons. The right-hand side hand-grip can be swivelled up to ±90°.
Monitoring is done on a 3” LCD panel of ‘approximately 1,030,000 pixels’ (about 64x43mm, implying
about 1240x830 pixels). It can be swivelled up or down nearly 90° and the image can be enlarged 2:1 for
focusing. Most of the camera control is via this display as a touch-screen, although the system controls are
via a menu system driven by buttons on the hand-grip. Since the touch-screen is capacitative, it responds
only to fingers, not finger-nails or styli. Also, the accessory mirror-box prevents access to the screen.
Canon XC10 July 2015
2
Tests have been conducted in line with EBU R.118. This document is a report of the results of
the tests defined in Tech3335 and is not an endorsement of the product.
EBU - Tech 3335 : Methods of measuring the imaging performance of
television cameras for the purposes of characterisation and setting
Alan Roberts, July 2015
SUPPLEMENT 17: Assessment of a Canon XC10 camera
Many of the menu items have little or no effect on image quality. Those that have significant effect are
highlighted. Control of the camera is split between ‘proper’ menus and item-by-item control from the touch
screen. In boxes with a range of numeric settings, the values indicate the range, and no scales are given.
Default settings, where known, are underlined. Only settings relevant for video shooting are listed here,
although many operate for video shooting, replaying and photo modes. All measurements were made using
Manual mode.
The tested camera was an E version which has items related to 50Hz, the U version (NTSC) has items related
to 59.94Hz and drop-frame timecode. Only the E version is covered here.
In the tables, items that have an important effect on picture appearance are highlighted with grey
background. Rather than making recommendations for settings, I have included measurement results from
which the user can make his own decisions.
This is not intended as a replacement for reading the manual.
1. Switches, Connectors and Menu settings
Switches and connectors
Left side Focus Manual/auto switch
Disp. (Assign 1) Cycles screen info
Push AF (Assign 2) Find auto focus
Mic 3.5mm stereo input
HDMI out Video, disabled if WiFi enabled
USB Computer connection
Handle top Start/Stop Traditional red button
Photo/Video Mode switch
Wheel Asignable rotary control
On/Off Power button
Play Playback start/stop
Shooting mode dial
Manual, Av (aperture priority), Tv (shutter priority), P (program AE),
Auto, Scn (special scene)
Handle back Magnify (Assign 3) LCD enlarger, x2
Menu button
Menu joystick
Handle right Headphone 3.5mm stereo output
Bottom back Memory card slots
Battery compartment
Menu settings Access by menu button or touch screen. Select/set via joystick or
touch or drag relevant part of the screen. Submenus are indented.
Camera setup
Item Mode Range Description Face detection & tracking MAvTvPScn On, Off Not tested
Auto slow shutter PAv On, Off Blurs in low light
ND filter All On, Off 3 stops, x1/8
Flicker reduction All Auto, Off Reduce lighting flicker
Auto ISO limit or AGC
limit AvTvP
Auto, ISO or
Auto, dB
Set max ISO160~4000 or
Gain 0dB ~ 28.5dB
Canon XC10 July 2015
3
Tests have been conducted in line with EBU R.118. This document is a report of the results of
the tests defined in Tech3335 and is not an endorsement of the product.
Shockless WB All On, Off
Onscreen markers All
Off, Level (white), Level
(grey), Grid (white), Grid
(grey)
Level is mid-high horizontal,
Grid is 3x3
Focus ring direction All Normal, Reverse
Focus ring response All Fast, Normal, Slow
ISO/Gain All ISO, Gain Changes some menu items
Gain increment All Normal, Fine
Recording setup
Item Mode Range Description 4k/HD All 4k clips, HD clips Also on screen, bottom left
4k recording All 25p/305Mb/s, 25p/206Mb/s
HD recording All 50p/50Mb/s, 50i/35Mb/s,
25p/35Mb/s
Slow & Fast motion1 All
4k Off, x2, x4, x10, x20, x60,
x120, x1200 Also on screen, bottom right
HD Off, x1/4, x1/4, x2, x4, x10,
x20, x60, x120, x1200
Available space in memory All Reports on cards
Initialize All CF, SD Format memory cards
Timecode mode All Preset, Regen
Timecode running mode All Recrun, Freerun
Initial timecode All hh:mm:ss:ff Set the timecode
User bit type All Setting, Time, Date
Color bars All Off, EBU, SMPTE Swap EBU for ARIB in the U
version
1kHz tone All -12dB, -18dB, -20dB, Off Tone with bars
File numbering All Reset, Continuous
Audio setup
Item Mode Range Description Headphone volume All 0~8~15
Notification sounds All 0~8~15 Beeps for self-timer etc
Built-in mic wind screen All Auto (high), Auto (low), Off
Custom audio only
Built-in mic att All Auto, On, Off
Built-in mic freq response All Normal, Boost LF, Low cut,
Boost MF, Boost HF+LF
Built-in mic directionality All Mono, Normal, Wide
Mic Att All Auto, On, Off
Only when external audio is
plugged in
Mic low cut All On, Off
Mic terminal input All Line, Mic
Audio limiter All On, Off
Audio compressor All High, Low, Off
Wi-Fi setup
Item Mode Range Description Browser remote All Off, On Must be Off to enable HDMI
Browser connection
settings
All Set camera ID and port
Smartphone connection All
Access point connection All
Display MAC address All
________________________________________________________________________________________________ 1 Slow & Fast recording is not quite what it seems, e.g. x1/4 means recording 4 times over-speed, and setting 50p at
50Mb/s the camera will run at 25p and recording is at 18Mb/s. Beware, check other settings such as shutter speed to
make sure that recordings will be sensible.
Canon XC10 July 2015
4
Tests have been conducted in line with EBU R.118. This document is a report of the results of
the tests defined in Tech3335 and is not an endorsement of the product.
System setup
Item Mode Range Description Language All For the menus (English) Lots of languages
Time zone/DST All Lots of cities (Paris) Also Summer time On, Off
Date/Time All
Date/Time All Enter date and time
Date format All YMD, MDY, DMY MDY in U version
24 hour All Select Changes time to am/pm or
24h
LCD brightness All Horizontal scale Shows greyscale to help
LCD backlight All H, M, L
Fan All Auto, On The fan is very quiet
Wireless remote control All On, Off
Tally lamp All On, Off
Auto power off All On, Off 5 minutes to auto power-off
Assign button 1
Assign button 2
Assign button 3
All
Disp, Push AF, Magnification,
Zebra, Peaking, Digital tele-
conv, Powered IS, ND
filter,Start/stop, Photo
Default is 1=Disp, 2=Push
AF, 3=Magnifcation
Control Dial All Iris, Shutter, ISO/Gain Rotary control near Start/Stop
Customize FUNC menu All
Aperture, Shutter, Gain/ISO,
White balance, Mic level,
Exposure lock, AE shift,
Zebra, Peaking, Powered IS,
Digital tele-conv, Focus,
Magnification
Add/remove items from the
FUNC menu, display top right
Batt info All Display battery info
HDMI timecode All On, Off
HDMI rec command All On, Off
HDMI status All Display only
Distance units All Metres, Feet
Backup menu settings All Save, Load Saves menu to SD card
GPS auto time setting All Off, Auto update
GPS information display All Display
Certification logo display All Display
Firmware All Display
Reset all All No, Yes Full reset
Touch screen controls Submenus are indented.
Item Mode Range Description
FUNC Vertical list, right-hand side
Iris MAv F/2.8~F/11 In ¼ stop steps. Max aperture
reduces with zoom
Shtr MTv ½~1/50~1/2000 Shutter, 4 steps per factor of 2
ISO/GAIN2 M
Look
1,2,4
ISO 160~20000ISO
ISO stops 3 steps per factor of
2, GAIN in 1.5dB steps
GAIN 0~42dB
Look
3,5
ISO 500~20000ISO
GAIN 9~42dB
WB MAvTvPScn
Automatic, Daylight, Shade,
Cloudy, Flourescent,
Fluorescent H, Tungsten, K,
User 1, User 2
White balance. K can be set
2000 to 15000. User setting
are auto
Mic level MAvTvPScn Auto, Manual Manual can be set 0~100
Exposure lock AvTvPScn Manual, Off Manual can be set -3~+3
stops in ¼ steps
________________________________________________________________________________________________ 2 Noise is good up to 6400ISO (31.5dB) but the picture softens significantly above 1600ISO (19.5dB)
Canon XC10 July 2015
5
Tests have been conducted in line with EBU R.118. This document is a report of the results of
the tests defined in Tech3335 and is not an endorsement of the product.
AE shift AvTvP -2~0~+2 Stops in ¼ steps
Zebra All Off, 70,100 70%/100%
Peaking All Peak, Peak off Make sharp edges red in LCD
Powered IS All On, Off Image stabiliser
Digital tele-conv3 All On, Off X2 lens magnification
Focus3 All Touch spot to focus
Magnification3 All Off, On Touch to select area
FUNC
Vertical list, left-hand side
Look/Scn4
MAvTvP
1 Standard, 2 EOS Std, 3
Wide DR, 4 Cinema EOS Std,
5 Canon log, User 1, User 2
2 is too vivid, 5 needs post
Scn
Portrait, Sports, Night, Snow,
Beach, sunset, lowlight,
Spotlight, Fireworks
Audio scene All
Standard, Music, Festival,
Speech, Meeting, Forest &
birds, Noise suppression,
Custom
Metering mode AvTvP Standard, Spotlight, Backlight
Self-timer5 All Off, 2 sec, 10 sec
Image stabilizer All Off, Standard, Dynamic
Pre Rec6 All On, Off 5 second pre-rec
Other FUNC controls
Item Mode Range Description HD/4k All 4k clips, HD clips Bottom left of screen
Slow & Fast motion1 All
4k Off, x2, x4, x10, x20, x60,
x120, x1200 Bottom right of screen
HD Off, x1/4, x1/4, x2, x4, x10,
x20, x60, x120, x1200
________________________________________________________________________________________________ 3 Item not available unless specifically added via ‘Customize FUNC menu’
4 Standard has good knee for general purpose shooting, EOS Std is too vivid, 3 4 and 5 have a more gentle knee for a
film-type look 5 Self-timer appears in the menu list but could not be enabled in the tested camera.
6 Pre-recording is available only in HD recording.
Canon XC10 July 2015
6
Tests have been conducted in line with EBU R.118. This document is a report of the results of
the tests defined in Tech3335 and is not an endorsement of the product.
2. Measurements
All measurements were made on frames captured onto a SDHC card for HD, CFast card for 4k. Images for
this document were extracted as BMP files or LZW-compressed TIF files using Edius 7.31 and Edius 8. I
shall use the EBU system of designating scanning standards. Live viewing was done on a 42” consumer
grade plasma Panasonic display with ‘studio’ settings. In this section,
2.1. Colour performance
A standard Colorchecker chart was exposed, using daylight. The camera was allowed to auto-white balance
and to auto-expose. Fig 1 shows the performance for the five standard ‘Looks’.
Look 1 (standard) is good for normal television, Look 2 (EOS vivid sharp and crisp) is quite vivid and over-
saturated, Look 3 (wide dynamic range) is desaturated and nice for a film look, Look 4 (EOS camera) is less
saturated but possibly a little too much, Look 5 (Canon Log, wide dynamic range for post-processing) is
under-exposed and under-saturated.
Choosing the right Look is a personal matter, but Look 1 is fine for normal TV, Look 3 is good for a film
look. The others are all acceptable but may bring problems with noise and/or saturation levels.
There was no response to infra-red, clearly there must be an IR-stop filter in the optical path. This augurs
well.
2.2. Resolution and aliasing
2.2.1. Resolution for 4K (UHD-1)
Tests were made at F/5.6. The usual zone plate test
chart was framed to fill exactly half the width and
height of the image. Thus the calibrated dimensions
should all be doubled.
Fig.2 shows one quadrant of the luma pattern which
now reaches the 3840x2160 limits of UHD-1, plus
one quadrant of the smaller pattern which now
reaches the 7680x4320 of UHD-2. The recording
was intra-frame at 305Mb/s, using Look 1.
The modulation is extinguished above 3000
horizontally, and 1700 vertically. There is fairly
strong diagonal aliasing which is inevitable with a
Bayer-patterned sensor. There is no evidence of any
aliasing above the limits of UHD-1, therefore either
the lens cannot resolve it or there is an optical low-
pass filter tailored to the limits of UHD-1.
Interestingly, the diagonal resolution exceeds both
horizontal and vertical extinction limits. This
implies that the Bayer-decoding is the limit, rather
than the optics, which in turn suggests that there is
probably no optical low-pass filter and it is the lens which is providing the spatial filtering.
(a) Look 1 (b) Look 2 (c) Look 3 (d) Look 4 ( e) Look 5
Figure 1 Colorchecker p/50
Figure 2 4k luma p/50
Canon XC10 July 2015
7
Tests have been conducted in line with EBU R.118. This document is a report of the results of
the tests defined in Tech3335 and is not an endorsement of the product.
Fig. 3 shows quadrants of red and green. The red pattern resolution is little different from the green, which
confirms that resolution is limited by the lens.
The resolution performance is acceptable for Tier 2, UHD-1, although 10-bit output is required for full
compliance and so this camera should strictly not be accepted for 4k broadcast. Aliasing levels are
acceptably low.
Fig. 4 shows how the resolution changes with the ‘Look’ setting. Clearly, there is a considerable drop with
Look 5, and the alias levels have dropped dramatically as well. However, the resolution in Look 5 only just
exceeds the limits of HD, and while it is likely that post-processing would restore much of the resolution loss
the alias level would probably also rise.
Figure 2 4k red and green, p/50
(a) Look 4 (b) Look 5
Figure 4 4k luma
Canon XC10 July 2015
8
Tests have been conducted in line with EBU R.118. This document is a report of the results of
the tests defined in Tech3335 and is not an endorsement of the product.
2.2.2. Resolution for HDTV
Exposures were made at F/5.2. The zone plate chart
was framed to exactly fill the image. The image for
1920x1080 HD comes from a 3840x2160 part of the
sensor, thus there seems to be no need to decode the
Bayer pattern since each of the red and blue sub-
patterns are 1920x1080, and the green sub-pattern
has two, spatially offset, patterns of 1920x1080.
However, this would cause serious aliasing since the
optical low-pass filter (if there is one) must be
tailored to the 4k performance of the camera.
Therefore we should expect to see the effects of
down-scaling in the HD performance, i.e. some
aliasing.
Fig. 5 shows luma quadrants in p/50 with sharpness
set to minimum, zero, and using ‘Look 1’, described
as ‘Standard image for shooting clips’. The main
quadrant reaches 1920x1080, 3840x2160 in the
smaller pattern. There is little or no aliasing. Vertical
resolution is significantly better than horizontal, the
extinctions appear to be at about 1700x1080. Since the photo-sites are square, there seems to be no reason
why resolution should not also be square, i.e. horizontal and vertical should be the same.
Fig. 6 shows red and green quadrants. Since the image is from the compressed video the red pattern should
not have the same resolution as the green. However, the difference is quite small but appears to be
symmetrical. Nevertheless, it is quite acceptable.
Fig 7 shows the luma part of the chart with detail set to the default value 3, and to maximum 7. The default
value seems a good compromise between aliasing and resolution, while the maximum setting is clearly not
acceptable for good HD. Note that the maximum setting also brings up aliasing beyond the limits of HD,
which will cause problems with motion in MPEG coding. This also shows that the optics do not block
frequencies beyond HD.
Figure 5 HD luma p/50, sharp=0
Figure 6 HD red and green p/50, sharp=0
Canon XC10 July 2015
9
Tests have been conducted in line with EBU R.118. This document is a report of the results of
the tests defined in Tech3335 and is not an endorsement of the product.
Ass for 4k shooting, the ‘Look’ setting affects the resolution. Fig.8 shows the same part of the chart with the
‘Look’ set to 4 (similar to Canon EOS cameras) and 5 (wide dynamic range and colours for post-processing).
The resolutions are clearly very different, Look 4 appears to be much sharper although there is an associated
risk of some aliasing.
Fig. 9 shows the resolution for interlaced HD, with standard sharpness and Looks 1 and 5. The re-sampling
process which generates the interlaced output does not filter out the vertical components sufficiently to avoid
serious aliasing.
(a) sharp=3 (b) sharp=7
Figure 7 HD luma p/50,
(a) Look 4 (b) Look 5
Figure 8 HD luma p/50, sharp=3
Canon XC10 July 2015
10
Tests have been conducted in line with EBU R.118. This document is a report of the results of
the tests defined in Tech3335 and is not an endorsement of the product.
2.3. Noise, Dynamic Range and Sensitivity
2.3.1. Noise, Dynamic Range and Sensitivity for HDTV
The camera was exposed to a 6-step grey scale, tungsten illuminated. Multiple exposures were taken to
explore the dynamic range at ISO800 (13.5dB Gain), shooting HD using Look1. Fig. 10 shows the result,
noise levels plotted vertically versus signal level. The solid line is a trend line through (or nearly through) the
luma points.
Conventionally, the noise level would be
expected to rise near black since the differential
gain applied by gamma correction affects the
noise level, but here it drops consistently. There
are several possible reasons, but it is hardly
worth exploring these since the user has no
control over the noise performance. The level
at 50% video is about -48dB which is the
qualifying level for EBU R.118 HD Tier 1.
However, since the noise level falls from this
level towards black, the pictures look rather
less noisy than these figures imply.
Next, the camera was exposed to achieve 50% luma level
over the full range of ISO settings. Exposure was
controlled using the shutter and iris. Fig. 11 shows the
result, plotting the noise level versus the ISO setting
expressed in dB, with 0 being ISO 160. The ‘normal’
curve for this test would be a linear slope, rising by 3dB in
noise level with each 6dB in gain. The deviation from this
ideal indicates that there is some signal processing,
probably noise reduction, going on which is one of the
(a) Look 1 (b) Look 5
Figure 9 HD luma i/25, sharp=3
-62
-60
-58
-56
-54
-52
-50
-48
-46
-44
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1
Figure 10 HD noise profile, p/50 Look 1, ISO 800
-51
-50
-49
-48
-47
-46
-45
-44
-43
-42
-41
0 6 12 18 24 30 36 42
Figure 11 HD noise vs gain setting at 50% video
Canon XC10 July 2015
11
Tests have been conducted in line with EBU R.118. This document is a report of the results of
the tests defined in Tech3335 and is not an endorsement of the product.
possible reasons for the unconventional shape in Fig. 9. Nevertheless, the noise level of -48dB is held up to
32dB gain, which is the equivalent of ISO 6400.
A more likely reason for the unusual noise levels is a change in resolution with ISO setting. Fig. 12 shows
this clearly. Resolution drops considerably at the higher speed settings, so speeds higher than about ISO1600
(19.5dB) gain) ought not to be regarded as full HD even though the noise level is acceptable.
For dynamic-range measurements, multiple exposures were made of a Colorchecker chart at ISO 400
(7.5dB) using Look1, aiming to record exposures at which the white patch was just clipped, and the black
patch only just discriminable. These extremes were met with exposure difference of 80:1. The reflectances of
the white and black patches are 90.01% and 3.13% respectively, a ratio of 28.76:1. Thus the total dynamic
range must be about 2,300:1 about 11¼ stops. The same data was used to explore the opto-electronic
(gamma) curve, which appears to be the standard ITU.709 curve up to about 50% output level then curving
gently through a knee into a lower slope and reaching clipping level at about 450% exposure level. Thus this
curve appears to be able to cope with about 2.12 stops of headroom.
For comparison, the same test was done using Look5 and ISO500 (10.5dB) which was expected to have a
much wider range. The exposure difference was again 80:1, again making a total dynamic range of about
2,300:1. This is a surprise, since it implies that the only advantage in using the Looks is in the distribution of
this, apparently fixed, dynamic range. Again, the same data can be used to explore the gamma curve, and
again the curve is very like the ITU.709 curve up to about 40% output, then curving through a knee to reach
about 450% exposure level, about 2.12 stops of headroom, the same as for Look 1. Fig.13shows the curves.
For sensitivity calculations, more multiple exposures were used to find the aperture setting at which the peak
white patch reached 100%, when illuminated at 2000 lux. The camera was set to ISO800 (13.5dB gain) and
Look1, 1/25 second exposure. The signal levels are not clipped at 100% video level and it was simple to find
an exposure which reached exactly 100%, at between F/6.7 and F/7.3. So the sensitivity must be about F/5 at
(a) ISO 400 (7.5dB) (b) ISO6400 (32.5dB) (c) ISO 20000 (42dB)
Figure 12 HD zone plate softening with ISO setting, i/25 Look 1
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
1.1
0 1 2 3 4 5
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
1.1
0 1 2 3 4 5
(a) Look 1 (b) Look 5
Figure 13 gamma curves
Canon XC10 July 2015
12
Tests have been conducted in line with EBU R.118. This document is a report of the results of
the tests defined in Tech3335 and is not an endorsement of the product.
ISO800 and 1/50 exposure. However, since the gamma curve of Look1 reaches 100% video at 2.4 times the
exposure at which a genuine ITU.709 gamma curve would reach 100% video, the aperture which would
cause 100% signal without this knee must be a factor of 1/2.4 less than measured, i.e. 1.55 stops less light, so
the lens aperture at which 100% video would be recorded without the knee must be about F/6.7 at ISO800.
2.3.2 Noise, Dynamic Range and Sensitivity for 4k
The camera was exposed twice to the 6-step
grey scale, at F/4 and F/11. This gives 12
points in a plot of noise versus signal level,
Fig.14.
The plot lines are trend lines through the
relevant points and thus are only approximate.
However, it is clear that the shapes are
different from those at HD, these curves are
much more like the conventional shape where
the noise level is roughly proportional to the
slope of the gamma curve. This implies that
the down-conversion process for deriving HD
pictures is acting as a filter rather than an
interpolator, which is highly desirable.
Next, the camera was exposed to the 6-step grey
scale again, using Look 1 and p/50. The ‘speed’
was set to 6 values from ISO 500 to 16,000. Fig.
15 shows the results. Clearly, as the speed rises,
the noise distribution rises further towards black,
but there is still an eventual down-turn near
black in each case.
Noise around mid-grey, the usual signal level for
categorisation of cameras, is much less affected.
Even at 16,000 ISO the mid-grey noise is
acceptable although the increase near black is
rather steep.
Fig. 16 shows the noise level at mid-grey versus
speed, expressed as gain in dB. The level is
better than -50dB (the target level for EBU
R.118 UHD Tiers 1 and 2) for all speeds below 36dB
gain. However, the rise in level near black (see Fig. 14)
makes the pictures look more noisy than this measure
would suggest when the camera is set to high
photographic speed. Nevertheless, the noise levels are
acceptable provided the photographic speed is set not
greater than about 10,000 ISO or 24dB gain.
Fig. 17 shows how the resolution changes with speed.
For this test, the zone-plate chart was framed to fill the
image, i.e. the dimensions of the larger quadrant at
1920x1080 and the smaller 3840x2160. It shows that the
higher speeds cannot be properly regarded as 4k
resolution, but are quite good for HD. So, shooting at up
to about 10,00 in 4k is a good way to get HD performance using down-conversion from the full frame in
post-processing. The noise levels at 20,000 are unacceptable.
-55
-54
-53
-52
-51
-50
-49
-48
-47
-46
-45
-44
-43
-42
-41
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
500
1000
2000
4000
8000
16000
Figure 15 4k noise profiles
-55
-54
-53
-52
-51
-50
-49
-48
0 6 12 18 24 30 36 42
Figure 16 4k noise level at mid-grey versus camera
gain in dB
-59
-58
-57
-56
-55
-54
-53
-52
-51
-50
-49
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1
Figure 14 4k noise profile, Look 5, ISO 800
Canon XC10 July 2015
13
Tests have been conducted in line with EBU R.118. This document is a report of the results of
the tests defined in Tech3335 and is not an endorsement of the product.
2.4. Motion effects
The sensor has a conventional ‘rolling shutter’
process for readout, and so differentially
distorts objects moving in the frame. Fig. 18
shows parts of two frames from a sequence of
a small desk fan, the blades rotating such that
they appear to be almost stationary. The
distortion is quite obvious, and cannot be
corrected. This level of distortion is not
unusual in cameras with a rolling-shutter.
2.5. Conclusion
The recording coder bit-rates qualify the camera for all Tiers in R.118, however, the bit-depth is problematic.
The 4k resolution is limited by the optics, and there is some diagonal aliasing. The camera cannot officially
qualify for R.118 UHD-1 because the recording is 8-bit, whereas R.118 requires 10-bit or greater. Also,
R.118 demands interchangeable lenses for UHD-1. Nevertheless the noise performance meets the criteria for
both Tiers of UHD-1, and the 4k mode is a good way to shoot for HD production provided extreme speed
settings are not used.
The HDTV resolution and aliasing is good enough for the camera to qualify for R.118 HD Tier 1, as is the
noise level. It is only the monitoring and connectivity requirements which are not really met. Thus its
performance meets Tier 1, and it easily meets Tier 2L. It could be used in Tier 1 situations provided the
monitoring and connectivity problems can be dealt with.
There is no infra-red response, implying the presence of an IR-stop filter. Motion portrayal is affected by the
sensor rolling-shutter.
Figure 18 motion effects
(a) ISO 400 (b) ISO 6,400 (c) ISO 20,000
Figure 17 4k resolution, Look 1, 50mm, F/5.6